Syllabic Writing and Syllable Structuregene/papers/syllabaries.pdf · Syllabic Writing and Syllable...

Syllabic Writing andSyllable Structure

Eugene BuckleyUniversity of Pennsylvania

SUNY at Stony Brook — 21 April 2006

“written language is a product oflinguistic awareness” (Aronoff 1985)

Writing Systems andLinguistic Structure

• Reflects the categories present in spokenlanguage– segment, syllable, word, etc.

• Embodies a theory of language– though also affected by cultural history

Traditional Typology of Writing

• Ideographic– not truly linguistic; semasiographic

• Logographic– most often “morphographic”

• Syllabic– or is it moraic?

• Alphabetic– segments, potentially just consonants

Moraic or Syllabic?“In addition to the syllable and thesegment, writing systems exist that makeuse of the constituents mora and rhyme,and the notion head.”

“[Pure] syllabaries are exceedingly rare.”

Poser (1992)

Moraic or Syllabic?“After all, it is not really the syllable, butrather the light syllable which recurs as aunit of representation in one script typeafter another.”

Ratcliffe (2001)

Against the Moraic Analysis

•CV systems may include non-moraicelements

•CV systems may fail to write allmoraic elements

•CV is not a moraic constituent•CV is really just a minimal syllable

CV as a constituent

The mora-linked onset

Moras in Syllable Structure

Are they Or are theyconstituents? weight units?

σ

µ µ

t a n

σ

µ µ

t a n

Constituency of CVIf onset links to head mora:

[ [ ta ]µ [ n ]µ ]σ

If onset links directly to syllable:

[ t [ a ]µ [ n ]µ ]σ

Other Constituencies

t [ a ]µ [ a ]µ [ n ]µ

t [ a ]µ [ an ]µ

[ tr [ a ]µ ]σ

[ t [ a ]µ ]σ

[ ta ]µ [ a ]µ [ n ]µ

[ ta ]µ [ an ]µ CV:C

[ [ tra ]µ ]σCCV

[ [ ta ]µ ]σCV

Japanese kana

The ultimate moraic example

Japanese kana as moraic“Although the kana scripts are often calledsyllabaries, they are in fact moraicsystems. Each symbol in the kana scriptsrepresents one mora. Most of these are CVsequences, but final /N/ or /Q/ count asseparate morae, and vowel length adds amora.”

Rogers (2005)

Basic signs in kanaNumber of signs = number of moras

• (C)V, the core syllables• N, the “moraic nasal”• vowel length

– special sign in katakana– use second V sign in hiragana

Mora matches in hiragana

さむらい おかあさん<sa-mu-ra-i> <o-ka-a-sa-N>[sa]µ [mu]µ [rai]µµ [o]µ [kā]µµ [san]µµ

samurai okāsan“warrior” “mother”

Mora matches in katakana

ネクタイ ジーンズ<ne-ku-ta-i> <zi-L-N-zu>[ne]µ [ku]µ [tai]µµ [jīn]µµµ [zu]µ

nekutai jīnzu“necktie” “jeans”

Derived signs in kanaGemination using “Q”• small version of /tu/ = [tsɯ]• here, sign = mora

Complex onset in [CyV]• <Ci> plus small version of <yV>• here, sign ≠ mora• similar for some borrowed sequences

Mora matches in hiragana

しき しっき<si-ki> <si-Q-ki>[ši]µ [ki]µ [šik]µµ [ki]µ

šiki šikki“command” “lacquerware”

Mora matches in katakana

ロケット ヨーロッパ<ro-ke-Q-to> <yo-L-ro-Q-pa>[ro]µ [ket]µµ [to]µ [yō]µµ [rop]µµ [pa]µ

roketto yōroppa“rocket” “Europe”

Mora mismatches in hiragana

ちょっと じゅみょう<ti-yo-Q-to> <zi-yu-mi-yo-u>[čot]µµ [to]µ [ju]µ [myō]µµ

čotto jumyō“a little” “life span”

Mora mismatches in katakana

キャッシュ ファミコン<ki-ya-Q-si-yu> <hu-a-mi-ko-N>[kyaš]µµ [šu]µ [fa]µ [mi]µ [kon]µµ

kyaššu fami kon“cash” “video game”

Interim Summary 1

• Japanese graphs write CV, whichusually corresponds to a mora.

• But parallel techniques are used towrite non-moraic differences.

• Other examples:– Linear B “onset splitting”– Mongolian ‘Phags-pa codas

Ethiopic Alphasyllabary

A systematic syllabary

Ethiopic Alphasyllabary

ä u i a e (ɨ) oh ሀ ሁ ሂ ሃ ሄ ህ ሆl ለ ሉ ሊ ላ ሌ ል ሎḥ ሐ ሑ ሒ ሓ ሔ ሕ ሖm መ ሙ ሚ ማ ሜ ም ሞ

Ethiopic Alphasyllabary

ä u i a e (ɨ) oʔ አ ኡ ኢ ኣ ኤ እ ኦb በ ቡ ቢ ባ ቤ ብ ቦg ገ ጉ ጊ ጋ ጌ ግ ጎd ደ ዱ ዲ ዳ ዴ ድ ዶ

ʔa bu ɡi da

Minimal words in TigrinyaOne heavy syllable: Two light syllables:

bun ‘coffee’ lomi ‘today’

σ

µ µ

b u n l o m i

σ

µ

σ

µ

Minimal words in Tigrinya

bu-n = ቡን lo-mi = ሎሢ

σ

µ µ

b u n l o m i

σ

µ

σ

µ

One sign per mora?

ሰልሚ ሰላም ሳሊማsälmi sälam salima‘muddy soil’ ‘peace’ ‘lightning source’

σ

µ µ

s ä l

σ

µ

m i

σ

µ

s ä

σ

µ

s a

σ

µ

l i

σ

µ

m a

σ

µ µ

l a m

Gemination in Tigrinya

ʔabo ‘oleander’ ʔabbo ‘father’

σ

µ µ

ʔ a b

σ

µ

o

σ

µ

σ

µ

ʔ a b o

Gemination in Tigrinya

ʔabo = ኣቦ ʔabbo = ኣቦ

σ

µ µ

ʔ a b

σ

µ

o

σ

µ

σ

µ

ʔ a b o

Sign ≠ Mora

ኣቦ ʔa-bo ʔabo ‘oleander’ ʔabbo ‘father’

not ኣብቦ ʔa-b-bo

በለመ sä-lä-mä sälämä ‘be calm’ sällämä ‘adorn’

not ሰልለመ sä-l-lä-mä

Sign ≠ Mora

ይሰብሩyɨ-sʌ-b-ru yɨ.sʌb.ru ‘they break’

ይስበሩyɨ-sɨ-bʌ-ru yɨ.sɨb.bʌ.ru ‘they are broken’

Interim Summary 2

• Tigrinya writes CV, a supposedmora, and coda consonants.

• But omits gemination, which isexactly a mora.

• Other examples:– Cypriot– Akkadian (optional)

Against CV as a mora

Phonological considerations

Moras as Headed Constituents

Both moras and syllables impose minimal sonoritythreshold on their heads; syllabic ≥ moraic.

Zec (1988, 1995)

Berber Serbian Kwakwalavowel vowel σ vowelliquid σ liquid liquidnasal nasal μ nasal

σ,μ obstruent μ obstruent obstruent

But ultimately this is orthogonal to onset linking.

Moraic licensingBella Coola obstruent clusters often not insyllables, which are rather simple CRV:C.

σ

µ µ

l a kk

µ

t

µ

cBagemihl (1991)

‘ten’

Bella Coola reduplication

‘do too much’yaɬ-yaɬk→[ yaɬ ] k

‘wink’p-ɬa-ɬa→p [ ɬa ]

‘tree bark’pɬt-kn -kn-ɬp→pɬt [ kn ]

‘to taste’qps-ta-ta→qps [ ta ]

‘mother’s-tan-tan-mts→s [ tan ]

‘bear-berry’mil-milixʷ-ɬp→[ mi ] [ lixʷ ]

Bella Coola reduplicationBagemihl: copying begins at the first syllable

σ

µ

t a

µ

p

µ

s

µ

q

µ

k

σ

µ µ

y a ɬ

σ

µ µ

y a ɬ

σ

µ

t a

Alternative syllabificationFricatives can head syllables, and stray consonants aremoraless “semisyllables”.

Cho and King (2003)

σ

µ

t ap

µ

sq

σ

µ

t a

σς

Unmarked ReduplicantsThe Emergence of the Unmarked (TETU) in Optimality Theoryprevents copying of fricative-headed syllables: *Peak/Obstruent >> Align-Prefix

Carlson (1997)

σ

µ

t ap

µ

sq

σ

µ

t a

σς σ

µ

t ap

µ

sq

σ

µ

p s

σς

Branching moras affect timing

/ʕi.nab.hum/ /ki.taa.bi/ /ki.taab.hum/‘their grape’ ‘my book’ ‘their book’

σ

µ µ

a b

σ

µ µ

a b

examples from Levantine Arabic

σ

µ µ

a

Arabic consonant durations

Vµ Cµ Vµ VCµ

Jordanian 88.4 ms 67.6 msSyrian 114.4 80.9Lebanese 81.2 66.9

Broselow, Huffman, Chen, and Hsieh (1995)Broselow, Chen, and Huffman (1997)

Arabic long-vowel durations

Vµ Vµ Vµ VCµ

Jordanian 161.0 ms 131.6 msSyrian 123.9 112.2Lebanese 114.2 97.8

Broselow, Huffman, Chen, and Hsieh (1995)Broselow, Chen, and Huffman (1997)

Crosslinguistic V durations

Vµ VµVCµ VµVµ

Arabic 1 1.6 1.9Malayalam 1 1.6 2.2Runyambo 1 1.4 2.2

Broselow, Chen, and Huffman (1997)Hubbard (1994, 1995)

Onset/RhymeZhùyīn fúhào or bōpōmōfō : onset (+ glide) + rhyme

<d-ai>ㄉㄞ/daj/

<b-ai>ㄅㄞ/baj/

ㄅㄧㄠㄅㄥㄉㄠ

<b-enɡ>/bəŋ/

<d-au>/daw/

<b-i-au>/bjaw/

Onset/Rhyme in SyllablesPoser includes onset/rhyme systems in his typology, but it’snot clear if this is consistent with his moraic representations.

σ

µ µ

t a n

σ

µ µ

t a n

morasdefinerhyme

morasstraddleonset/rhyme

Interim Summary 3

• The evidence for CV as a moraicconstituent mora is not persuasive.

• The overall consensus in the field isfor onsets linked directly to thesyllable node.

• The “moraic writing system” analysisis not available.

CV as a syllable

An efficient corewith supplementation

Core syllables

• Every language has CV syllables• Some also have V syllables• (C)V is the minimal syllable

– “core”– “universal”– “maximally unmarked”

• Cf. Onset, NoCoda, *Complex

“Core”, not “Light”Tagalog recent perfective:

ka-ta-trabaho ‘work’ka-bo-bloaut ‘give special treat’

If relevant notion were light syllable, wouldexpect CCV instead.

Japanese syllables

Basic (C)V with five deviations:• Coda N• Coda C as part of geminate• Long vowel• Complex onset• Novel CV in borrowings

Supplemented Japanese syllables

+ small kanano new moranovel CV

+ small kanano new moraCyV

+ kana+ moraCVV

+ small kana+ moraCVC

+ kana+ moraCVN

WrittenSpoken

Japanese kana and moras

• By the nature of Japanese phonology,nearly every deviation from CV yields anew mora.

• But CyV and borrowed CV sequences arealso written with two kana.

• No good correlation between addition ofkana and addition of mora.

Mycenaean and Linear B• Core CV as in typical syllabary.• Complex onsets written with echo vowel.

– though not /s/ before a stop

• Codas not written.– cf. special case of final /ks/ and sonority

• No indication of vowel or consonant length.– except /w/ as second half of diphthong

• Small number of “optional” signs.

Unwritten Segments in Linear B

‘seed’<pe-ma>/spér.ma/

‘town’<wa-tu>/wás.tu/

‘around’<a-pi>/am.pʰí/

‘chief’<qa-si-re-u>/gʷa.si.léws/

‘shepherd’<po-me>/poj.mēn/

‘bronze’<ka-ko>/kʰal.kós/

Onset Clusters in Linear B

‘sheepfold’<ta-to-mo>/sta.tʰmós/

‘Knossos’<ko-no-so>/knōs.sós/

‘tripod’<ti-ri-po>/trí.pos/

‘of lands’<ko-to-na-o>/ktoj.nā.hōn/

‘leaɡue’<a-to-mo>/ar.tʰmós/

‘with bands’<de-so-mo>/de.smõjs/

Final “Onset” Clusters

‘axles’<a-ko-so-ne>/á.kso.nes/

‘Ethiopian’<ai-ti-jo-qo>/aj.tʰí.okʷs/

‘king’<wa-na-ka>/wá.naks/

‘two swords’<qi-si-pe-e>/kʷsí.pʰe.he/

‘spear points’<ai-ka-sa-ma>/aj.ksmáns/

Supplemented Linear B syllables

+ sign+ moraCVks

+ sign+ moraCVw

+ signno new moraCCV

no sign+ moraCV: , CVj

no sign+ moraCVC

WrittenSpoken

Linear B signs and moras• By the nature of Greek phonology, only some

deviations from CV yield a new mora.• The scribal practice largely ignored coda

consonants and other moraic elements, but didwrite clusters with non-falling sonority.– salience of onset due to C in CV ?

• Terrible correlation between addition of signand addition of mora.

Complex syllables in Cypriot

‘seed’<se-pe-re-ma>/spér.ma/

‘for’<ka-re>/gár/

‘man’<a-to-ro-po-se>/án.tʰrō.pos/

‘tablet’<ta-la-to-ne>/dál.ton/

‘for Perseutas’<pe-re-se-u-ta-i>/per.séw.tāj/

Similar to Linear B (and related to it), but writes most codaconsonants with an echo vowel or dummy <e> word-finally.

Supplemented Cypriot syllables

no sign+ moraCVCi.Ci

+ sign+ moraCVV

+ signno new moraCCV

no sign+ moraCV:

+ sign+ moraCVC

WrittenSpoken

Cypriot signs and moras• The essential CV system hasn’t changed.• Still deviates from a “moraic” system:

• Nonwriting of long vowels and geminates.• Writing of onset clusters.

• Supplementary strategies have simplybecome more plentiful.

• No useful correlation between addition ofsign and addition of mora.

Intentional minimality in Maya

‘house’tana<ta>

‘fish’kay<ka>

‘tree sap’pom<po>

‘net’paw<pa>

For signs whose origin is known with some confidence, wecan observe truncation of the morphographic value to CV.

Complex syllables in Maya

‘acquire shape’<pa-ta>/pat/‘coati’<či-ku>/čihk/

‘bird’<mu-ti>/mūt/

‘bone’<ba-ki>/bāk/

‘play ball’<pi-tsi>/pits/

‘dog’<tsu-lu>/tsul/

Writes coda consonants with a “synharmonic” vowel — or ifdisharmonic, indicates a complex vowel with length or /h/.

Supplemented Maya syllables

+ mora?CVhC+ sign CVk

+ moraCV:C

+ sign CVino new mora?CVC

WrittenSpoken

The extra signs serve not to mark moraic information, butrather to partly indicate linguistically important distinctions:a coda consonant, and vowel length or /h/.

Syllabic signs in AkkadianBasic

VCVVCCVC

Combined

CV-V = CV:CV-VC = CVCCV-V-VC = CV:C

All vowels and consonants are written, but vowellength and gemination are often omitted.

Complex syllables in Akkadian

‘selections’<liq-ti>/liq.tī/

‘clever’<nak-la>/nak.la/

‘pertains’<ba-aš-mu>/baš.mu/

‘whatever’<ma-la>/ma.la/

‘tablets’<ṭup-pa-a-ni>/ṭup.pā.ni/

‘I checked’<as-niq>/as.niq/

Akkadian moraicity/ra.bîm/ ‘of the big one’

<ra-bi-im><ra-bi-i-im>

/i.kaš.šas.si/ ‘defeats her’

<i-ka-ša-si><i-ka-aš-ša-si><i-ka-ša-as-si><i-ka-aš-ša-as-si>

Supplemented Akkadian syllables

1 in CV1 moraCV

1 in CVC2 in CV-VC3 in CV-V-VC

3 moras (?)CV:C

1 in CV2 in CV-V2 morasCV:

1 in CVC2 in CV-VC2 morasCVC

GraphsMoras

Akkadian mappings• Every sign is a possible syllable.• Combinations of signs extend the

system:– CV-VC for convenience or by necessity– CV-V for greater explicitness

• Priority to write every segment; moraicdistinctions secondary.– relevance of segments to all languages, even

if written with a syllabary

Korean syllable blocksAlphabetic han’gŭl defines “syllable” on a partlymorphological basis.

닭이 값을<talk-i> <kaps-ŭl>[tal.gi] [kap.sŭl]

‘chicken (SUBJ)’ ‘price (OBJ)’

Similar abstraction from surface syllabificationoccurs in syllabaries.

The notion “syllabary”

Poser (1992):“each syllable is represented by adistinct graph.”

Sproat (2000):“ought to imply that every syllableof the language has a graphemicsymbol associated with it.”

The notion “alphabet”

Poser (1992):“each segment is represented by adistinct graph.”

Sproat (2000):“symbols available to representevery phonemic segment of thelanguage.”

Alphabets in practiceIdeally, every segment has a graph.

[p] → p[i] → e

But in many cases, digraphs or otherstrategies are used to represent certainsegments.

[č] → ch[ŋ] → nɡ

Syllabaries in practiceIdeally, every syllable has a graph.

[ta] → た[ko] → こ

But in many cases, digraphs or otherstrategies are used to represent manysyllables.

[tan] → たん[kyo] → きょ

Efficiency in Japanese

46Kana(with diacritics = 25 more)

276Syllables(CVX, some CV excluded)

20Phonemes(14 C, 5 V, plus N)

Practical implementation• An alphabet can reasonably reach full

coverage of the segments.– but of course often doesn’t

• A syllabary must be much larger to coverall the possible syllables of a language.– so this situation is rarely achieved

• Yi, perhaps Chinese– for simple reasons of scale, not because the

syllable isn’t really the relevant unit

Implications

Adapted from Poser (1992)

1. Correspondence betweenWriting and Speech

“Writing systems make use of essentiallythe same constituents that phonologistsdo.”

Yes; but this set of constituents does notinclude the mora.

2. Phonological Constituency

“Their role in writing systems providesadditional support for the syllable, mora,segment, and rhyme.”

Well, not the mora.

And we have to remember that writingsystems are the products of culturalhistory, not pure reflections of language.

3. Alleged Atomicity of the Syllable

“The extreme rarity of syllabaries togetherwith the evidence for subsyllabic analysis ...eviscerates the arguments for thepsychological atomicity of the syllable thathave been based on the putativepredominance of syllabaries.”

This conclusion remains true, even ifsyllabaries are common; that constituent issalient and important, but not atomic.

Thanks!

Gene BuckleyDept of Linguistics619 Williams HallUniversity of PennsylvaniaPhiladelphia, PA 19104-6305

gene@ling.upenn.edu